This randomized, multinational, multicenter study was designed to determine the response rate of gemcitabine monotherapy and cisplatin/etoposide combination therapy in chemotherapy-naive patients with advanced, recurrent, and/or metastatic non-small cell lung cancer (stage IIIA [if inoperable], IIIB, or IV). One group of patients received gemcitabine 1,000 mg/m2 intravenously once a week for 3 weeks (days 1, 8, and 15) followed by a 1-week rest period. The second group received cisplatin 100 mg/m2 intravenously on day 1 of each 28-day cycle in combination with etoposide 100 mg/m2, administered on days 1, 2, and 3 following the cisplatin infusion. Each patient was allowed to remain on study up to a maximum of six cycles. The planned interim analysis was based on the 117 patients in the study, 116 of whom were randomized up until November 20, 1995. The efficacy analysis was performed on the 107 patients who had data from a minimum of two cycles, whereas the safety analysis was based on data from all 116 randomized patients. In the gemcitabine arm there were 10 of 52 (19%) partial responders; in the cisplatin/etoposide arm there were four (7%) of 54 partial responders. There was a statistically significant difference in the response rates between the two arms, with a 95% confidence interval of 0.6% to 32.1% (P = .040). The median time to progressive disease was 4.2 months for gemcitabine patients and 3.7 months for cisplatin/etoposide patients. There was significantly more alopecia and nausea and vomiting in the cisplatin/etoposide arm compared with the gemcitabine arm, as well as two cases of neutropenic sepsis in the cisplatin/etoposide arm. These data indicate that single-agent gemcitabine is at least as effective as the combination of cisplatin/etoposide in the treatment of advanced non-small cell lung cancer and has an improved safety profile.